#1177 Fixed unroll syntax for old drivers.

This commit is contained in:
XMRig 2019-09-18 23:46:51 +07:00
parent e3fcb99d84
commit 133cd30b2e
5 changed files with 2744 additions and 2744 deletions

View file

@ -604,7 +604,7 @@ __kernel void hashAes1Rx4(__global const void* input, __global void* hash, uint
__local const uint* const t2 = ((sub & 1) == 0) ? (T + 512) : (T + 1536);
__local const uint* const t3 = ((sub & 1) == 0) ? (T + 768) : (T + 1280);
#pragma unroll(8)
#pragma unroll 8
for (uint i = 0; i < inputSize / sizeof(uint4); i += 4, p += 4)
{
uint k[4], y[4];

View file

@ -72,7 +72,7 @@ __kernel void fillAes_name(__global void* state, __global void* out, uint batch_
const __local uint* const t2 = (sub & 1) ? (T + 512) : (T + 1536);
const __local uint* const t3 = (sub & 1) ? (T + 768) : (T + 1280);
#pragma unroll(unroll_factor)
#pragma unroll unroll_factor
for (uint i = 0; i < outputSize / sizeof(uint4); i += 4, p += 4)
{
uint y[4];

File diff suppressed because it is too large Load diff

View file

@ -902,7 +902,7 @@ __global uint* generate_jit_code(__global uint2* e, __global uint2* p0, __global
{
int prefetch_data_count;
#pragma unroll(1)
#pragma unroll 1
for (int pass = 0; pass < 2; ++pass)
{
#if RANDOMX_PROGRAM_SIZE > 256
@ -929,7 +929,7 @@ __global uint* generate_jit_code(__global uint2* e, __global uint2* p0, __global
prefetch_data_count = 0;
#pragma unroll(1)
#pragma unroll 1
for (uint i = 0; i < RANDOMX_PROGRAM_SIZE; ++i)
{
// Clean flags
@ -1314,7 +1314,7 @@ __global uint* generate_jit_code(__global uint2* e, __global uint2* p0, __global
// Sort p0
uint prev = p0[0].x;
#pragma unroll(1)
#pragma unroll 1
for (int j = 1; j < prefetch_data_count; ++j)
{
uint2 cur = p0[j];
@ -1344,7 +1344,7 @@ __global uint* generate_jit_code(__global uint2* e, __global uint2* p0, __global
__global int* prefetched_vgprs = prefecth_vgprs_stack + num_prefetch_vgprs;
#pragma unroll(8)
#pragma unroll 8
for (int i = 0; i < RANDOMX_PROGRAM_SIZE; ++i)
prefetched_vgprs[i] = 0;
@ -1359,7 +1359,7 @@ __global uint* generate_jit_code(__global uint2* e, __global uint2* p0, __global
const uint size_limit = (COMPILED_PROGRAM_SIZE - 200) / sizeof(uint);
__global uint* start_p = p;
#pragma unroll(1)
#pragma unroll 1
for (int i = 0; i < RANDOMX_PROGRAM_SIZE; ++i)
{
const uint2 inst = e[i];

View file

@ -1658,7 +1658,7 @@ uint32_t inner_loop(
const int32_t sub2 = sub >> 1;
imm_buf[IMM_INDEX_COUNT + 1] = fprc;
#pragma unroll(1)
#pragma unroll 1
for (int32_t ip = 0; ip < program_length;)
{
imm_buf[IMM_INDEX_COUNT] = ip;
@ -1934,7 +1934,7 @@ __kernel void execute_vm(__global void* vm_states, __global void* rounding, __gl
const uint32_t workers_mask = ((1 << WORKERS_PER_HASH) - 1) << ((get_local_id(0) / IDX_WIDTH) * IDX_WIDTH);
const uint32_t fp_workers_mask = 3 << (((sub >> 1) << 1) + (get_local_id(0) / IDX_WIDTH) * IDX_WIDTH);
#pragma unroll(1)
#pragma unroll 1
for (int ic = 0; ic < num_iterations; ++ic)
{
__local uint64_t *r;